Scaling down and low‐frequency noise in MOSFET’s. Are the RTS’s the ultimate components of the 1/f noise?

A large amount of experimental results show that for MOSFET’s with channel area exceeding 10 μm2, 1/f noise is clearly present, while for transistors of area less than 1 μm2, Random Telegraph Signals (RTS) are emerging, giving rise to lorentzian spectra.This is rather well explained by the carrier number fluctuation model recently developed in our laboratory. A relation is established between the level of 1/f noise and the defectiveness of the device. An important result concerns the reliability of integrated circuits: beyond a certain limit of scaling down, one can show that it is impossible to predict, with a given accuracy, the low frequency noise of an individual transistor.A critical question also arises: are the RTS’s the ultimate components of the 1/f noise? In this paper, we emphasize the interest of observing extensively intermediate cases MOSFET’s (i.e., with a surface of a few μm2). In these cases, several traps are involved, leading to multi level temporal signals. In the presented experiments...